Received: 25 March 2016
Revised: 13 May 2016
Accepted: 30 May 2016
DOI 10.1111/pedi.12406
REVIEW ARTICLE
Medication-induced diabetes mellitus David R. Repaske MD, PhD Department of Pediatrics, University of Virginia, Charlottesville Corresponding Author: David R. Repaske, Department of Pediatrics, University of Virginia, Charlottesville, VA (
[email protected]).
Epidemiological studies and case reports have demonstrated an increased rate of development of diabetes mellitus consequent to taking diverse types of medication. This review explores this evidence linking these medications and development of diabetes and presents postulated mechanisms by which the medications might cause diabetes. Some medications are associated with a reduction in insulin production, some with reduction in insulin sensitivity, and some appear to be associated with both reduction in insulin production and insulin sensitivity. KEYWORDS
diabetes mellitus, drug-related side effects and adverse reactions, insulin resistance, insulinsecreting cells, metabolic side effects of drugs
1 | I N T RO D UC T I O N A variety of medications have been associated with development of
2 | COMMON MEDICATIONS THAT DECREASE BOTH INSULIN SENSITIVITY AND INSULIN RELEASE
diabetes. Establishing a precise cause and effect relationship between a medication and development of diabetes is challenging
2.1 | Glucocorticoids
for several reasons. Side effects of most medications are rare and clinical studies of medications typically evaluate effectiveness and
Glucocorticoids are widely used medications and clinical experience
are not powered to evaluate side effects. Diabetes is a common dis-
indicates that diabetes is a frequent side effect. Ingle first reported
ease and there is always a question of whether it would have devel-
glucosuria in rats receiving glucocorticoid.1 A human retrospective
oped if the person had not taken the medication in question.
epidemiological study of adult Medicaid patients carried out over
Patients are often taking multiple medications and so it is hard to
9 years identified almost 12 000 persons with new onset diabetes
determine which medication was responsible for the side effect. For
and closely matched controls which did not develop diabetes.
example, organ transplant patients may be taking a glucocorticoid
Patients who were taking glucocorticoid were more likely to be in the
and a calcineurin inhibitor, so an epidemiological study has difficulty
diabetes group (relative risk 2.23) and the relative risk was dose
ascribing the side effect to one or the other or the combination of
dependent.2 One intuitive hypothesis for the mechanism by which
these medications. There are also covariants such as the weight gain
glucocorticoids cause diabetes is development of insulin resistance.
associated with steroids or antipsychotics making it difficult to know
To test this hypothesis, hyperinsulinemic-euglycemic clamps with
if development of diabetes was a primary or secondary effect of the
tritiated glucose tracer were performed on 10 healthy subjects
medication. When we turn to deciphering the mechanism by which
after 7 d of placebo and after 7 d of prednisolone.3 This study
a medication caused diabetes, there are often multiple plausible
demonstrated that prednisolone induced insulin resistance in liver,
molecular mechanisms identified in animal or in vitro studies, but
muscle, and adipose tissue by decreasing the ability to suppress
these do not necessarily identify the clinically relevant mechanism(s)
hepatic gluconeogenesis, as well as by decreasing peripheral glucose
in humans. Finally, for pediatrics, there are few studies in children
disposal in muscle and adipose tissue. There are multiple plausible
and data from adults can only be assumed to apply to children.
biochemical mechanisms for development of insulin resistance in the
A summary of the most common medications that have been associ-
postreceptor insulin-signaling pathways in all tissues (reviewed by
ated with development of diabetes and their most plausible mech-
van Raalte4). In addition, there is evidence for decreased insulin-
anism is presented in Table 1.
induced vasodilation in muscle leading to reduced glucose delivery to
392
© 2016 The Authors. Pediatric Diabetes published by John Wiley & Sons Ltd.
wileyonlinelibrary.com/journal/pedi
Pediatric Diabetes September 2016; 17: 392–397
This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made.
393
REPASKE
TABLE 1
Medications associated with development of diabetes and their most plausible mechanisms Insulin resistance and insulin deficiency
neurotransmitters that interact with D2 and 5-HT2a receptors in the brain and also a spectrum of other receptors that vary from one antipsychotic to another, including 5-HT1a, 5-HT2c, α1, M1, M3, and H1
Insulin deficiency
Atypical antipsychotics Beta blockers
Insulin resistance
receptors. The use of atypicals began in 1989 and case reports noted
Beta-adrenergic agonists
weight gain and new diabetes.11 Onset of diabetes in some cases
Glucocorticoids
Calcineurin inhibitors Growth hormone
preceded weight gain. Clinical experience in adults indicates a rela-
Nicotinic acid
Diazoxide
tively high association between clozapine and olanzapine and new
Protease inhibitors (first gen)
Didanosine
Statins
Diphenylhydantoin
Megestrol
onset diabetes.12 Risperidone and quetiapine have a lower association, and aripiprazole and ziprasidone are least associated with diabetes. An epidemiological study from Veterans Administration (VA) data
Gatifloxicin
in 1999-2001 evaluated patients who had never had diabetes and
L-asparaginase
were newly started on an antipsychotic.13 Due to the VA formulary
Pentamidine
in effect at the time, only haloperidol, olanzapine, risperidone, and
Thiazide diuretics
quetiapine were studied. Patients initiated on haloperidol were used as the control group and the increased hazard ratio (HR) for develop-
muscle beds, reducing the opportunity for muscle to clear glucose 4
ment of diabetes on each of the other medications was evaluated. There was an increased risk (HR ~1.6) for each of the three atypicals
from the blood.
Glucocorticoids also have additional direct effects on release of
that was not significantly different among the three. Interestingly, the
insulin from beta cells.5 Six healthy subjects were given three mixed
greatest risk was in the youngest age group (74 years) where the HR was about 3 for each medica-
on day 2 and without any additional prednisolone on day 3. Plasma
tion. Another study followed 82 clinic-based patients without diabe-
glucose was elevated during the day 2 challenge, but c-peptide was
tes who were newly started on clozapine and followed for 5 years for
unchanged from baseline suggesting that the beta cells were unable
development of diabetes.14 The study was complicated by an average
to release additional insulin to overcome the peripheral insulin resist-
weight gain of 525 g per month. After 5 years, 36% of the patients
ance resulting from the acute exposure to glucocorticoid. There are
had developed diabetes. This is a dramatic number of patients with
also a number of plausible postreceptor biochemical targets in the
new diabetes, but lack of a control group and the fact that these sub-
beta cell for glucocorticoid suppression of insulin secretion.4
jects also gained an average of 500 g per month complicate the inter-
The initial common sense approach to glucocorticoid-induced dia-
pretation of results.
betes is to minimize the glucocorticoid dose as much as possible. There
In pediatric patients, a retrospective study by Bobo et al of nearly
are few studies of optimal medical treatment for glucocorticoid-
29 000 Medicaid patients (age 6-24 years) initiating atypical antipsy-
induced diabetes.6,7 Guidelines are largely based on expert opinion and
chotic therapy were compared with very carefully matched controls
therefore vary. The consensus is that postprandial excursions in blood
who were initiating a different type of neuroactive medication such
glucose are more significant than fasting hyperglycemia in most
as lithium, antidepressant, clonidine, or stimulant. Development of
instances, and therefore recommendations are made for use of glinides,
diabetes was evaluated over the following 4 years.15 Approximately
thiazolidinediones (TZDs), glucagon-like peptide 1 (GLP-1) analogs, or
0.2% of the patients in the control group developed diabetes com-
dipeptidyl peptidase 4 (DPP-4) inhibitors.6–10 The TZD, troglitazone,
pared to 1.1% of the patients in the atypical antipsychotic group.
has been shown to improve postprandial glucose excursions (lower
Another national Medicaid data analysis by Rubin et al of nearly
area under the glucose curve in an oral glucose tolerance test (OGTT)),
110 000 10-18 year olds who initiated atypicals compared to
decrease HbA1c, decrease low-density lipoprotein, and increase insulin
1 200 000 less closely matched controls that were seen for a mental
secretion in patients on chronic glucocorticoid,9 but is not widely
health concern but not started on atypicals,16 showed a smaller over-
recommended due to side effects of weight gain, heart failure, and
all association with diabetes than the previously cited study.15 In the
bone fractures, all of which may be additive to similar side effects of
20 mo of this study, 0.25% of the controls developed new diabetes
glucocorticoids in the USA, the drug was removed from the market in
while 0.38% of the patients started on atypicals had developed diabe-
2000. There are few studies of metformin in glucocorticoid diabetes,
tes. The HR was therefore 1.51. However, looking back at the 20 mo
and it has not generally been recommended as many patients taking
time point in the Bobo study, similar results were seen in that about
glucocorticoid have renal failure or may be prone to lactic acidosis,
0.15% of the control and 0.4% of the atypical using group had devel-
which constitute contraindications to metformin.8,10 Consensus is that
oped diabetes. This is because the rate of diabetes development in
blood glucose consistently over 200 mg/dL associated with glucocorti-
controls decreased with time in the Bobo study, while the rate of
8
coid therapy should be treated with insulin.
development of diabetes in the atypical group continued to increase at a near constant rate. In the Rubin study,16 there was an increased
2.2 | Atypical antipsychotic medications
risk for development of diabetes on ziprasidone or aripiprazole of about 1.6 compared to the risk of development of diabetes on risper-
Atypical antipsychotics are also clinically associated with develop-
idone. There was not a statistically significant difference in risk for
ment of diabetes. This is complicated by the frequently associated
diabetes development between quetiapine, olanzapine, or risperi-
weight gain. The atypical antipsychotics (“atypicals”) antagonize
done. These pediatric data are different from what is typically seen in
394
REPASKE
adults where ziprasidone and aripiprazole have reduced risk of devel-
consensus expert opinion is to approach atypical antipsychotic-
opment of diabetes and olanzapine confers a greater risk.
associated diabetes as for type 2 diabetes.
Mechanisms whereby atypical antipsychotics might cause diabetes must include weight gain and consequent insulin resistance. This concept is demonstrated in a study of 48 normal subjects who were randomized to a 15-17 d course of risperidone, olanzapine, or placebo.17 A hyperglycemic clamp was performed before and after taking the
3 | O T H E R M E D I C A T I O N S T H A T D E C RE A S E BOTH INSULIN SENSITIVITY AND INSULIN RELEASE
medication. There was no decrease in insulin release in response to the medication, but there was a decrease in insulin sensitivity that was
There are several additional diabetogenic medications that appear to
entirely accounted for by the acute increase in weight of about 3 kg in
have mechanistic targets in both the pancreas and the periphery. The
the antipsychotic group that did not occur in the placebo group.
first generation HIV protease inhibitors impair first phase insulin
Other studies in rats have found evidence for decreased insulin
secretion by decreasing activity of voltage-dependent (distinct from
release from beta cells in vitro in the presence of clozapine.18 At high
the ATP-sensitive) K-channels in beta cells thereby limiting calcium
glucose concentration, clozapine reversibly hyperpolarizes rat beta
influx and insulin secretion. They also have a postinsulin receptor
cells and decreases insulin secretion. Another in vitro study used rat
action in muscles, decreasing translocation of GLUT4 glucose trans-
islets incubated in physiologic glucose in the presence of carbachol,
porters to the muscle cell membrane and thereby limiting the capac-
an M3 agonist that mimics vagus nerve stimulated-release of insulin
ity of muscle to take glucose out of the bloodstream.28
19
from islets.
Clozapine and olanzapine (which have M3 antagonist
Statins have an association with new onset diabetes and carry a
activity) blocked this “neural mediated” insulin release, but risperi-
Food and Drug Administration warning to this effect. Patients with
done, ziprasidone, and haloperidol (which do not have M3 antagonist
baseline impaired glucose tolerance who were started on statins had
activity) did not.
a statistically significant HR of 1.32 for development of diabetes.29
In mice, 5-HT2c receptor activation in the brain has an insulin-
Several meta-analyses of statin-associated diabetes show an odds
like effect by activating MC4R neurons that signal to the periphery to
ratio of about 1.1 for development of diabetes when taking statin for
both reduce insulin resistance by activation of protein kinase B
up to 5 years compared to placebo.30 Hypotheses about the mechan-
(PKB)/Akt in muscle and reduce glucose production by decreasing
ism include inhibition of insulin release via attenuated glucose signal-
activity of two key enzymes in the gluconeogenesis pathway, phos-
ing in the beta cell and/or increased beta cell apoptosis and
phoenylpyruvate carboxykinase (PEPCK) and glucose-6-phosphatase
decreased insulin sensitivity at muscle through decreased expression
in liver.20 Zhou proposes that atypicals that are antagonists at 5-
of the glucose transporter, GLUT-4. Statins may also induce insulin
HT2c receptors may block these neural insulin-like effects in humans
resistance by suppression of adiponectin. There is evidence that the
and thereby contribute to hyperglycemia.20
magnitude of these effects vary from statin to statin, but each has
These studies suggest that some of the increase in diabetes asso-
been associated with a small increase in diabetes risk.30
ciated with atypical antipsychotics is due to associated weight gain,
The lipid-lowing drug, nicotinic acid, has long been known to
but that there appear to be other plausible mechanisms for the asso-
raise blood glucose and HbA1c in patients with diabetes.31 In mice, it
ciation, including reduced neural signaling from the brain to the
acts through a specific G-protein-coupled receptor, GPR109a, which
periphery that results in decreased insulin release and decreased neu-
triggers insulin resistance in obese mice, by a still undefined mechan-
ral insulin-like effects. Studies cited above suggest that there may be
ism, and reduces insulin secretion from beta cells.32,33
a stronger association between atypical antipsychotics and diabetes in the young and in the old, and the atypicals that are most closely associated with development of diabetes in adults may be somewhat different from those associated with diabetes in children.
4 | A CO M M O N M E D I C A TI O N T H A T D E C R E A S E S I N S U LI N R E L E A S E
Therapy for diabetes associated with atypical antipsychotics may start with discontinuing or changing to an alternate antipsychotic, if possible. A number of small studies in children and in adults have
4.1 | Calcineurin inhibitors
demonstrated effectiveness of metformin in limiting weight gain with
Calcineurin inhibitors are immunosuppressants that are widely used
atypical antipsychotics.21,22 Metformin has been demonstrated to
post-transplant to block NF-AT (nuclear factor of activated t cells)
23
prevent weight gain when used early in therapy with atypicals,
or
induction of interleukin-2 production in lymphocytes. Tacrolimus, a
promote weight loss in patients who had already gained weight on
macrolide antibiotic, and cyclosporin A, a cyclic polypeptide, are struc-
therapy with atypicals.24,25 A few studies also demonstrate improve-
turally dissimilar calcineurin inhibitors. Tacrolimus is the more potent
ment in metabolic parameters such as fasting glucose, but have not
but clinically is associated with more side effects. Cyclosporin A has
directly examined treatment of established diabetes associated with
cosmetic side effects of hirsutism and gum hyperplasia. A study of
atypical antipsychotics. Some of the studies include lifestyle modifica-
over 11 000 renal transplant patients from 1999 to 2000 found that
tion alone and in combination with metformin and they demonstrate
15% of patients on any calcineurin inhibitor developed diabetes in the
additive effects of lifestyle and of metformin on weight manage-
6 mo after transplant.34 The incidence of new diabetes among these
21,26,27
ment.
However, maintaining lifestyle modification is particu-
same patients was 20% for those on tacrolimus, suggesting that it is
larly challenging in patients on antipsychotic medications. The
associated with higher risk of development of new diabetes. This
395
REPASKE
higher risk persisted for the 3 years that patients were followed in this
the HR for development of diabetes on a thiazide is 2.07 and the risk
study with over 32% of patients on tacrolimus and 22% of patients on
is related to the serum potassium with each 0.5 mEq/L decrease in
any other calcineurin inhibitor developing diabetes. This study was
serum potassium increasing the risk of diabetes by 45%.41
complicated by cotherapy with glucocorticoids. To investigate the
Didanosine, an antiretroviral reverse transcriptase inhibitor nucle-
mechanism of cyclosporin-associated diabetes, nine patients who
oside agonist, can trigger pancreatitis in up to 5%-10% of patients
were awaiting renal transplant and were not taking glucocorticoid
and some will have consequent beta cell injury.42 In addition,
were studied before and after 2 wk of cyclosporin administration
didanosine-associated hypokalemia has been postulated to reduce
using hyperglycemic clamps.35 Glucose was maintained at 160 mg/dL
insulin secretion in the absence of pancreatitis.
and second phase (120-180 minutes), but not first phase (
